Infection is one of the leading causes of death in hematology due to the need for large amounts of repeated use of hormones, immunosuppressive drugs, broad-spectrum antibacterial drugs, chemotherapeutic drugs, etc. However, there is no meta-analysis of the diagnostic efficacy of mNGS for infection in hematology patients. We included 17 studies with 18 groups to perform an evaluation of the diagnostic value of mNGS for infection in hematology patients. In our meta-analysis, mNGS showed superior diagnostic efficacy. The pooled sensitivity and specificity of mNGS for infectious diseases in hematology patients were 89.6% (95% CI: 88–91) and 56% (95% CI: 44–69), respectively, indicating an excellent diagnostic performance of mNGS for infection in hematology patients. Our results are similar to the data of a retrospective study on the diagnostic performance of mNGS for infection in hematologic patients, in which the sensitivity of mNGS for pathogens was 82.6% and the specificity was 59.0% [29]. Moreover, the pooled positive likelihood ratio, negative likelihood ratio, and diagnostic odds ratio were 2.10 (95% CI: 1.48–3.30), 0.22 (95% CI: 0.15–0.32), and 10.29 (95% CI: 4.61–19.72), respectively. The AUC of the SORC curve was 0.88 (95% CI: 0.85–0.90), indicating that mNGS has a high diagnostic value.
The vast majority of the studies we included were premedicated with antibiotics, which have previously been reported to affect the positivity of CMTs. Compared to blood cultures, mNGS is less affected by widely empirical antibiotics [17, 30]. mNGS targets nucleic acid fragment sequences of pathogens that survive longer in plasma or other tissue fluids. Therefore, mNGS still has satisfactory positive rates despite the use of broad-spectrum antibiotics. The studies we included described that the overall positive rate of pathogens detected by mNGS was significantly higher than that of CMTs (80.21% vs 25.00%, P < 0.001; 72.6% vs. 31.4%, P < 0.001; 43% vs. 14%, P < 0.001, 68.37% vs 37.76%, P < 0.001) [17, 20, 28, 29]. All of the above findings support that mNGS remains relatively advantageous for patients with previous antibiotic exposure.
The included studies had moderate heterogeneity in sensitivity (I2 = 43%, P = 0.03) using a fixed-effects model and substantial heterogeneity in specificity (I2 = 95%, P<0.01) using a random-effects model. We further explored the sources of heterogeneity by subgroup analysis. For sensitivity, the appropriate samples for mNGS may contribute to the heterogeneity. The sensitivity of the subgroup of samples selected according to the type of infection was higher than that of the other group (0.911 vs. 0.832, P = 0.02). Several published guidelines recommend selecting the appropriate specimen for mNGS based on the type of infection [31]. Therefore, in clinical practice, the most appropriate specimen should be evaluated and selected by the clinician. The heterogeneity of specificity was high (I2 = 95%, P < 0.01). We speculate that there are several reasons. First, subgroup analysis showed that research type contributes to the heterogeneity of specificity. Prospective studies can choose the proper proportion of uninfected and infected populations in their inclusion, while retrospective studies included mostly infected patients and only a minority of noninfected patients. Second, some of the studies chose CMTs as the gold standard, which has a high false negative rate, leading to a low specificity [15–17, 20, 21, 25]. This is also evidenced by our subgroup analysis of the gold standard (0.718 VS. 0.349, P < 0.01). Due to the high price of mNGS in clinical practice, it is often used as a complementary tool to CMT for failing to detect pathogens, so we are more interested in the sensitivity rather than specificity of mNGS.
In the real world, the application of mNGS is still a complementary diagnostic examination after the failure of traditional tests to identify the pathogen. Therefore, more prospective studies should be conducted in the future to explore the sensitivity and specificity of mNGS in the diagnosis of infection in hematology patients. On the other hand, considering that the difficulty of detecting pathogens is different for different types of infection, future studies should also explore the diagnostic value of mNGS in different types of infection, such as bloodstream infection, pulmonary infection, and urinary tract infection and other infection.
Currently, mNGS is emerging as a technology that plays an important role in the detection of pathogens in infectious diseases. The advantages of mNGS are faster, more comprehensive and more accurate data analysis, especially in the detection of specific pathogens [7, 32]. Our present study demonstrates the value and potential of mNGS in clinical practice in hematology patients. Our subgroup analysis illustrates the importance of selecting the appropriate specimens to be sent for mNGS in patients who are to undergo mNGS and that selecting the most appropriate specimens according to the site of infection helps to improve the detection of pathogens by mNGS. In our study, subgroup analysis showed that the population of patients (adult/pediatric vs. pediatric) and the method of mNGS sequencing (DNA vs. DNA/RNA) were not the cause of the source of heterogeneity. At the same time, mNGS did not show a significant advantage in patients undergoing hematopoietic stem cell or neutropenia, which indicates that mNGS will not be more beneficial in patients with HSCT or neutropenia.
We have shortcomings in this study. The first is that most of the studies we included were retrospective studies rather than prospective clinical controlled studies (RCTs), which can have some bias. Second, some of the studies had relatively small sample sizes and were not convincing enough to detect the diagnostic efficacy of mNGS. Third, there are several other sources of heterogeneity in our pooled specificity, which complicates the interpretation of the results of the pooled specificity. Because of the differences in genome length and sequencing platforms of different types of microorganisms, it is impossible to establish a uniform positive standard for all microorganisms [33–35], so we did not perform subgroup analysis for positive criteria of mNGS. However, different positive criteria may affect the diagnostic efficacy of mNGS. In the future, an updated guideline on mNGS positivity criteria for clinical practice and how to interpret mNGS result reports is needed.
Our analysis revealed the clinical utility of mNGS for infection in hematology patients. To better exploit its value, clinicians need to select the appropriate samples to send for mNGS according to the type of infection. At present, mNGS has some disadvantages, such as not being widely available in clinical practice because of its expensive price, and the criteria for positivity and interpretation of mNGS are not uniform. In the future, mNGS will be an excellent tool to diagnose infection in hematology patients and help with treatment.